UNCLASSIFIED UNCLASSIFIED A Modified Constant-Stress Coupon for Enhanced Natural Crack Start during Fatigue Testing Witold Waldman, Robert Kaye and Xiaobo Yu Aerospace Division Defence Science and Technology Group DST-Group-TR-3252 ABSTRACT This report details the development of a modified constant-stress coupon for use in fatigue testing. This novel coupon design has a significantly greater surface area along the notch boundary that is subjected to the peak stress, and it is useful in studies of the probabilistic growth of in-service fatigue cracks where fatigue life is governed by the most severe defect. The extended region of uniform stress is achieved by shape optimisation of the notch boundary, resulting in the elimination of the highly-localised stress concentration that is a characteristic feature of a traditional dog-bone coupon. The presence of an extensive region of uniform stress increases the incidence of fatigue cracking from small naturally-occurring surface imperfections or discontinuities, as well as more uniformly distributing the fatigue cracking over the region of constant stress. Stress intensity factors have also been computed for simulated crack growth trajectories for edge cracks starting at various locations distributed along the notch boundary. Use of the constant-stress coupon reduces the number of coupons that need to be tested in order to attain desired statistical confidence levels, leading to significant time savings and greatly reduced costs in conducting fatigue testing programs studying the initiation and growth behaviour of small cracks. RELEASE LIMITATION Approved for public release
111
Embed
A Modified Constant-Stress Coupon for Enhanced Natural ... · A Modified Constant-Stress Coupon for Enhanced Natural Crack Start during Fatigue Testing Executive Summary The Structural
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
UNCLASSIFIED
UNCLASSIFIED
A Modified Constant-Stress Coupon for Enhanced Natural Crack Start during
Fatigue Testing
Witold Waldman, Robert Kaye and Xiaobo Yu
Aerospace Division Defence Science and Technology Group
DST-Group-TR-3252
ABSTRACT
This report details the development of a modified constant-stress coupon for use in fatigue testing. This novel coupon design has a significantly greater surface area along the notch boundary that is subjected to the peak stress, and it is useful in studies of the probabilistic growth of in-service fatigue cracks where fatigue life is governed by the most severe defect. The extended region of uniform stress is achieved by shape optimisation of the notch boundary, resulting in the elimination of the highly-localised stress concentration that is a characteristic feature of a traditional dog-bone coupon. The presence of an extensive region of uniform stress increases the incidence of fatigue cracking from small naturally-occurring surface imperfections or discontinuities, as well as more uniformly distributing the fatigue cracking over the region of constant stress. Stress intensity factors have also been computed for simulated crack growth trajectories for edge cracks starting at various locations distributed along the notch boundary. Use of the constant-stress coupon reduces the number of coupons that need to be tested in order to attain desired statistical confidence levels, leading to significant time savings and greatly reduced costs in conducting fatigue testing programs studying the initiation and growth behaviour of small cracks.
A Modified Constant-Stress Coupon for Enhanced Natural Crack Start during Fatigue Testing
Executive Summary
The Structural and Damage Mechanics Group in the Airframe Technology and Safety Branch of Aerospace Division has been deeply involved in research to develop and apply structural and damage mechanics methods and technologies to enhance the safety, availability, and reduce the cost of ownership of airframes in service with the Royal Australian Air Force. Included amongst its wide range of capabilities are a unique and novel shape optimisation technology for the repair and fatigue life enhancement of airframe components, advanced computational and analytical tools and methods for improved determination of fatigue and crack growth life of airframes, and integrated analytical and numerical modelling of both complete and local structures for the improved assessment of structural response and capability.
This report details the development of a modified constant-stress coupon for use in fatigue testing. A constant-stress coupon is a novel design that has a significantly greater surface area along the notch boundary that is subjected to the peak stress. This extensive region of uniform stress is achieved by shape optimisation of the boundary of the notch, resulting in the elimination of the highly-localised stress concentration that is a characteristic feature of a traditional dog-bone fatigue testing coupon. The shape optimisation method that has been used here is based on the advanced and powerful shape optimisation technique that has been pioneered by the Structural and Damage Mechanics Group, and which has recently been modified and ported to be used with the Abaqus finite element analysis software code.
As the constant-stress fatigue coupon design is being used in studies involving crack growth, an assessment of its performance from a fracture mechanics perspective was also conducted. This involved the detailed computation of crack-growth trajectories and stress intensity factors for through-thickness edge cracks emanating from points distributed along the shape-optimised notch boundary. An advanced two-dimensional boundary element analysis technique was used for these calculations. The results reported here will serve to permit comparison with the crack paths that occur during the fatigue testing of the coupons.
UNCLASSIFIED
UNCLASSIFIED
The use of constant-stress fatigue coupons will significantly assist research into improving the fundamental understanding of the probabilistic growth of in-service fatigue cracks where the fatigue life is governed by the most severe defect. The presence of an extensive region of uniform stress will increase the incidence of fatigue cracking from small naturally-occurring surface imperfections or discontinuities, as well as potentially more uniformly distributing the fatigue cracking over the region of uniform stress. Use of the constant-stress coupon will reduce the number of coupons that need to be tested in order to attain desired statistical confidence levels, leading to significant time savings and greatly reduced costs in conducting fatigue testing programs to study the initiation and growth behaviour of small cracks. The outcomes of this research into fatigue cracking will be applied to the long-term structural integrity management of both ageing and new ADF airframes.
UNCLASSIFIED
UNCLASSIFIED
Authors Witold Waldman Aerospace Division Mr Witold Waldman completed a BEng (with distinction) in Aeronautical Engineering at the Royal Melbourne Institute of Technology in 1981. He commenced work in Structures Division in 1982, at what was then the Aeronautical Research Laboratory. He has published a number of papers and reports, and his experience has focussed on stress analysis using finite element and boundary element methods, structural mechanics, fracture mechanics, computational unsteady aerodynamics, structural dynamics testing, digital filtering of flight test data, nonlinear optimisation, and spectral analysis. His recent work has been in the areas of structural shape optimisation and the computation of stress intensity factors. He is currently a Senior Research Engineer in the Structural and Damage Mechanics Group in the Airframe Technology and Safety Branch of Aerospace Division within the Defence Science and Technology Group.
Robert Kaye Aerospace Division Mr Robert Kaye joined what was then the Structures Division of the Aeronautical Research Laboratory in 1990 as a structural engineer with a background in full-scale testing. The first three years at DSTO were spent in evaluation of bonded repairs primarily using finite element methods. Included in that was the analysis of repairs to fuselage skin lap-joints, wing skin planks and bulkhead frames. More recently, he has been involved with structural and mechanical aspects of full-scale fatigue test installations. In particular, he played a key role in the development of a low-stiffness air-spring for the application of static load to a vibrating airframe. This work was followed by a period of several years doing research and development into the alleviation of stress concentrations by way of adaptive shape optimisation. This has been applied to concave metallic free boundaries and to the adhesive layer and end tapering of boron patches bonded to metallic structure. Upon his retirement, he is now a DST Honorary Fellow in Aerospace Division.
Xiaobo Yu Aerospace Division Dr Xiaobo Yu completed a BEng in Naval Architecture Engineering in 1988, and a MEng in Structural Mechanics in 1991, both from Shanghai Jiao Tong University; and a PhD in Civil Engineering from the University of Sydney. He has held a lecturing position in Shanghai Jiao Tong University and several senior research positions in the University of Sydney and the University of NSW, and was also attached to Pacific Engineering Systems International as a CRC-ACS Visiting Research Fellow. In 2007 he joined what was then the Defence Science and Technology Organisation, and he is currently Science Team Leader for Structural Mechanics in Aerospace Division. In the context of airframe safety and life assessment and life extension, his main interests and responsibilities include structural shape optimisation, computational structural analysis and multi-axial fatigue.
2. STRESS ANALYSIS OF TYPICAL DOG-BONE COUPONS WITH NON-OPTIMAL NOTCH PROFILES ................................................................................................... 3 2.1 Notch with large 100-mm constant-radius arcs .................................................................. 3 2.2 Notch with constant central width and large constant 100-mm radius fillets ............... 4 2.3 Notch with constant central width and moderate constant 30-mm radius
4. DESIGN OF THE MODIFIED COUPON WITH OPTIMAL CONSTANT-STRESS NOTCH PROFILE .......................................................................................................... 7 4.1 Nominal geometry and material properties ....................................................................... 7 4.2 Initial finite element mesh ..................................................................................................... 8 4.3 Performing the shape optimisation ...................................................................................... 8 4.4 Raw results of shape optimisation........................................................................................ 8 4.5 Preparation of raw optimal shape for manufacture ........................................................... 9
5. 2D FRACTURE MECHANICS ASSESSMENT OF MODIFIED CONSTANT-STRESS COUPON ....................................................................................................................... 12 5.1 Analysis of uncracked coupon ............................................................................................ 12 5.2 Computation of crack trajectories for edge-cracked coupon .......................................... 13 5.3 Computation of Beta factors for through-thickness edge-cracked coupon .................. 13 5.4 Computation of Beta factors for other crack geometries ................................................. 14
APPENDIX A: FADD2D INPUT FILE FOR UNCRACKED NOTCH COMPOSED OF LARGE 100-MM CONSTANT-RADIUS ARCS ..................................... 35
APPENDIX B: FADD2D INPUT FILE FOR UNCRACKED NOTCH WITH CONSTANT CENTRAL WIDTH AND LARGE 100-MM RADIUS FILLETS ................. 36
APPENDIX C: FADD2D INPUT FILE FOR UNCRACKED NOTCH WITH CONSTANT CENTRAL WIDTH AND MEDIUM 30-MM RADIUS FILLETS .............. 38
APPENDIX D: FORTRAN 90 SOURCE CODE FOR THE CREATEDOGBONECOUPONMODEL PROGRAM ........................................................... 40
APPENDIX E: INPUT DECK USED TO CREATE FADD2D MODEL OF ORIGINAL CONSTANT-STRESS COUPON USING CREATEDOGBONECOUPONMODEL PROGRAM ........................................................... 49
APPENDIX F: FADD2D INPUT FILE FOR MODIFIED CONSTANT-STRESS COUPON WITH ONE CENTRALLY-LOCATED THROUGH-THICKNESS EDGE CRACK .............................................................................................................................. 50
APPENDIX G: IGES FILE OF COORDINATES FOR MODIFIED CONSTANT-STRESS COUPON ....................................................................................................................... 65
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED
Nomenclature a crack length E Young’s modulus of the material F boundary correction factor (beta factor) K stress intensity factor Kt net-section stress concentration factor Ktg gross-section stress concentration factor S local tangential stress at notch surface S∞ remote uniaxial tension stress x x-coordinate y y-coordinate σ stress along notch boundary σmax maximum tangential stress υ Poisson’s ratio of the material
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 1
1. Introduction Prior work in DST Group by McDonald (1999a) has involved the computation of stress concentration factors (SCFs) for various fatigue test coupon profiles composed of straight lines and circular arcs. These coupons were required for use in fatigue testing to aid in the fatigue life management of RAAF airframes. To help maximise test machine usage efficiency and to provide the necessary statistical data for accurate estimation of fatigue life, it was desirable to test many short cracks on a single specimen. To achieve this, a test coupon was required that had a notch with a relatively large surface area of uniform stress (where the stress is constant within a specified small range). This led McDonald (1999b) to establish the design of an aluminium coupon that maximised the length of uniform stress under uniaxial tensile loading.
Following on from the work of Mattheck et al. (1992), the Structural and Damage Mechanics Group within DST Group has successfully developed and employed shape optimisation technology for the reduction of peak stresses occurring at typical stress concentrations such as fillets and holes (Heller, Kaye and Rose 1999; Waldman, Heller and Chen 2001; Burchill and Heller 2004; Heller et al. 2009, Waldman 2015). In essence, the method relies on adding material where stresses are high and removing material where stresses are low. Waldman and Heller (2006, 2015) have also significantly extended this in-house shape optimisation capability in order to be able to address problems that involve the simultaneous reduction of multiple stress peaks. The implementation of the method for use with the MSC Nastran finite element analysis (FEA) code that was developed by Braemar (2005) has also recently been ported by Kaye and Waldman (2016) to make use of the Abaqus FEA code running on a Linux–based computer system.
A distinct characteristic of shape-optimal designs is the presence of one or more large regions of uniform stress along the stress concentrator boundary, where the stress is constant within a small tolerance. Hence, in DST Group, Westcott (2010) and Kaye (2010) have previously taken advantage of this feature by applying shape optimisation methods to design specialised fatigue test coupons that have extensive regions of uniform stress. These shape-optimised coupon geometries are characterised by having the desired extensive regions of nearly-constant stress along a large proportion of the notch boundary, with the peak stress in the coupon corresponding to the stress level that is associated with the region of uniform stress.
The advantage offered by a constant-stress fatigue coupon is that there is a much larger area along the notch surface that is subjected to the peak stress, as the highly-localised stress concentration produced by a traditional dog-bone coupon is eliminated. It is therefore anticipated that the incidence of fatigue cracking from small naturally-occurring surface imperfections or discontinuities will be duly increased, as well as potentially being more uniformly distributed over the region of constant stress. This can serve to reduce the number of coupons that need to be tested, resulting in significant time savings and greatly reduced costs in conducting fatigue testing programs. The desired outcome from such
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 2
fatigue tests is to improve the fundamental understanding of the probabilistic growth of fatigue cracks from small naturally-occurring defects, where the in-service fatigue life of the component is assumed to be governed by the most severe defect that is present (Cetin, Härkegård and Naess 2013). In addition, Molent, Barter and Wanhill (2011) and Barter, Molent and Wanhill (2012) have noted that surface imperfections and discontinuities are features that are inherent to the material and the manner of production of the component, and that the existence of such defects in typical aircraft alloys leads to fatigue cracking in tests of both coupons and components.
As a result of the novel properties offered by a constant-stress fatigue test coupon, such designs have recently begun to be extensively utilised by the Aircraft Forensic and Metallic Technologies Group within DST Group (Shekhter et al. 2015; Loader, Shekhter and Turk 2015; Loader et al. 2016; Niclis and Harrison 2016; Shekhter et al. 2016; Turk 2016a; Turk 2016b; Turk and Niclis 2016). Prior to this, in conducting their fatigue test program on multiple coupons that have a low SCF (low-Kt), Yu et al. (2014) have noted that, over a region of low stress concentration, the lead crack that dictates the fatigue life is often the fastest growing crack that starts from one of the worst discontinuities. They referred to this phenomenon as “natural crack start”, which cannot be reproduced using traditional dog-bone coupons such as those defined in ASTM (2012), because the stress concentration that is inherent in the design of those coupons is significant enough to dictate the starting positions of the fatigue cracks. Yu et al. (2014) have also noted a further limitation, in that, by confining the crack starting locations to the area of higher stress, fewer initial discontinuities and local material properties are sampled, which affects the growth of small cracks. As a result, the largest crack that eventually results in failure may start from a less severe discontinuity or grow at a slower rate. Also, when the largest crack starts from a location that is next to, but not exactly at, the peak stress location, the use of nominal peak stress overestimates the actual fatigue driving force. Consequently, Yu et al. (2014) have concluded that the fatigue test results obtained from traditional types of dog-bone coupons will exhibit more scatter in total fatigue life, and any distribution of life data would be skewed to a longer life, when compared to those on an actual aircraft structure.
The effectiveness of the prior optimal coupon design has been confirmed by Yu et al. (2014) through experimental strain measurement as well as fatigue tests under spectrum loading. However, as designed, these coupons have on occasion experienced undesirable premature failures in the grip region during other subsequent test programs. These failures typically occurred while undertaking testing at low load levels, with commensurately long specimen fatigue lives. A significant factor contributing to these failures is considered to be the use of oversized test specimen grips, which affect the gripping action. This is thought to produce fretting damage, which then leads to instances of fatigue failure in the region where the coupon is being clamped by the grips. Therefore, it became apparent that a need exists for the design of a Modified Constant-Stress Coupon with a view to reducing the incidence of coupon failures originating in the grip area. The performance of the coupon whose design is reported on herein is expected to be better because of the increased width of the new coupon in the grip area. It must also be kept in
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 3
mind that the previous fatigue testing work undertaken with the narrower specimen has already produced an extensive usable body of results. In order to maintain compatibility with those results, it is desirable that the length of the constant-stress region in the modified coupon be very similar to that of the original design. This Report documents the design of this new Modified Constant-Stress Coupon.
In order to provide some background that serves as a basis for comparison, Section 2 gives the results of analyses of a variety of typical non-optimal non-constant stress dog-bone coupon designs. One of those designs has in fact previously been used in a coupon fatigue testing program in the DST Group. Section 3 then goes on to provide the results of a stress analysis of the original constant-stress coupon design. The design of the new Modified Constant-Stress Coupon is then described in Section 4. The results of a detailed fracture mechanics analysis of potential cracking along the notch boundary of the new coupon are presented in Section 5, and it includes simulated crack-growth trajectories and associated stress intensity factors for through-thickness edge cracks of various lengths. Finally, the conclusion is presented in Section 6.
2. Stress analysis of typical dog-bone coupons with non-optimal notch profiles
It is considered useful to develop a better quantitative understanding of the stress distribution in the notch region of some typical dog-bone axial fatigue test coupon geometries. For this purpose, a number of two-dimensional ¼-symmetry boundary element models of coupon designs are created here and analysed using the FADD2D boundary element code (Chang and Mear 1995; Newman et al. 2006). The assumed linear-elastic material properties for these aluminium coupons are a Young’s Modulus of E = 72.4 GPa and a Poisson’s ratio of υ = 0.33. Each model is subjected to a uniform uniaxial tensile load of 100 MPa applied to the ends of the specimen, and this was deemed to be capable of providing stress results in the notch region (the area of interest) comparable to those that would be expected if the actual grip loading could be more accurately simulated. Plane stress conditions were assumed in the analysis, which was deemed to be reasonable as the thickness of these coupons is small relative to the other dimensions.
2.1 Notch with large 100-mm constant-radius arcs
In prior DST Group work reported much earlier by McDonald (1999a), a simple notch design was developed that utilised a circular arc of large constant radius to produce a notch profile. The purpose behind this design was to try to minimise the stress concentration effect associated with the notch. This type of notch design is shown in Figure 1, and the hatched areas represent the grip length of 40 mm that had been adopted. The coupon dimensions for this notch design were: notch radius = 100 mm, total coupon
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 4
length = 160 mm, width = 40 mm, thickness = 6.35 mm, notch half-length = 38 mm, notch depth = 7.5 mm, and net section half-width = 12.5 mm.
A FADD2D ¼-symmetry model was created and analysed using the previously specified linear elastic material properties for aluminium (see Appendix A for a listing of the input deck, which is also located in Objective folder fAV1044714). The FADD2D-computed normalised stress response along the notch boundary, σ/σmax, is shown in Figure 2. Here σ is the tangential stress, and σmax is the maximum tangential stress. The net-section SCF is Kt = 1.081 for this design, which compares very well with Kt = 1.082 obtained from a 3D analysis using the StressCheck finite element code (McDonald 1999a). The gross-section SCF is Ktg = 1.730. The stress response obtained here consists of a broad peak, centred on x = 0 mm. These results show that the stress changes by 8% over approximately 30 mm of the notch length (–15 mm ≤ x ≤ +15 mm), and by 33% over approximately 60 mm of the notch length (–30 mm ≤ x ≤ +30 mm). Hence, it is evident that the constant-radius specimen design cannot meet the requirement of having a significant region with uniform stress.
2.2 Notch with constant central width and large constant 100-mm radius fillets
An alternative dog-bone coupon notch design that is composed of a long constant-width centre section with transitioning fillets of large constant 100-mm radius is shown in Figure 3. This dog-bone coupon geometry was selected on the basis that it is similar to those that are used in many fatigue testing programs, often with circular through-thickness holes added to the centre section. The geometry shown in Figure 3 is expected to result in a considerable constant-stress zone in the central constant-width region. However, it is also anticipated that a significant stress concentration will occur in the fillet regions.
A ¼-symmetry 2D model of the geometry shown in Figure 3 was created and analysed using the FADD2D boundary element program (see Appendix B for a listing of the input deck, which is also located in Objective folder fAV1044714). The previously specified linear elastic material properties for aluminium were used, and the applied load once again consisted of a uniform uniaxial tensile load of 100 MPa. The computed normalised stress response along the notch boundary, σ/σmax, is shown in Figure 4. The net-section SCF for this coupon design is Kt = 1.071, while the gross-section SCF is Ktg = 1.714. It is noted that the stress response has a broad peak, located at about x = ±23.9 mm. The stress is quite constant over approximately 30 mm of the notch length in the central region of the coupon (–15 mm ≤ x ≤ +15 mm), albeit at a value of about σ/σmax = 0.93, which is 7% lower than the peak stress. The left and right peaks (from symmetry) in the stress distribution along the notch boundary occupy about 40 mm of the total notch length (–35 mm ≤ x ≤ –15 mm and +15 mm ≤ x ≤ +35 mm). Although this design has a considerable zone of constant stress in its central region, the presence of the significant peak in the stress distribution would be expected to result in crack growth in that particular area during fatigue testing, rather than in the zone of constant stress.
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 5
2.3 Notch with constant central width and moderate constant 30-mm radius fillets
Another dog-bone coupon notch design, this time composed of a constant-width centre section with transitioning fillets of moderate constant 30-mm radius, is shown in Figure 5. These fillets have a much smaller radius than the 100-mm radius fillets used in the previous section and, as a result, they are expected to produce a more severe stress concentration effect.
A FADD2D ¼-symmetry 2D model of this geometry was created and analysed using the previously specified linear elastic material properties for aluminium (see Appendix C for a listing of the input deck, which is also located in Objective folder fAV1044714). The applied load once again consisted of a uniform uniaxial pressure of 100 MPa. The computed normalised stress response along the notch boundary, σ/σmax, is shown in Figure 6. The net-section SCF for this design is Kt = 1.240, while the gross-section SCF is Ktg = 1.984. These values are considerably more severe than those obtained when the 100-mm radius fillets were used. The stress response has a broad peak, located at about x = ±23.5 mm. The stress is quite constant over approximately 30 mm of the notch length in the central region of the coupon (–15 mm ≤ x ≤ +15 mm), albeit at a value of about σ/σmax = 0.80, which is 20% lower than the peak stress. The left and right peaks (from symmetry) in the stress distribution occupy about 40 mm of the total notch length (–31 mm ≤ x ≤ –15 mm and +15 mm ≤ x ≤ +31 mm). Although this design has a considerable zone of constant stress in its central region, the presence of the significant peak in the stress distribution would be expected to result in crack growth originating somewhere in that particular area during fatigue testing, rather than in the zone of constant stress.
3. Prior optimal constant-stress notch profile Some years ago now, the original DST Group shape optimisation code was ported to work with the MSC Patran and Nastran finite element analysis software, and the specifics of this version of the code are covered by Braemar (2005). The process used for preparing an optimal shape for manufacture is described by Wescott and Heller (2009). Taken in combination, these procedures were used by Wescott, Jones and Heller (2010) and then Kaye (2010) to design the original constant-stress fatigue test coupon.
The first constant-stress specimen was designed by Wescott, Jones and Heller (2010) using 2D shape optimisation techniques, as the then implementation of the code was not working correctly for 3D open boundary problems. Their specimen was 6.35 mm thick and had a nominal grip area of 40×40 mm at each end. The grip load was modelled as a constant traction (tangential pressure) during the optimisation process, and was implemented using nodal forces. A short 5 mm buffer zone between the grip line and the end of the notched section was included. As the specimen is 160 mm in total length, this
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 6
results in a notch length of 70 mm. The minimum section width was 24 mm at the centre of the specimen, corresponding to a notch depth of 8 mm. Minimum radius of curvature constraints in the range 5 mm to 25 mm were investigated, and a value of 20 mm was chosen for the production version of this design. The net-section Kt was 1.03 and the length of notch profile with up to 2% stress variation was 50 mm.
After correcting a bug in the code, Kaye (2010) subsequently ran a 3D version of the shape optimisation algorithm, which for the unsmoothed optimal notch shape produced a net-section Kt of 1.03. This specimen, which is shown in Figure 7, differed slightly from the earlier one, in that the depth of the notch was 7.5 mm rather than 8 mm. The total arc length of the constant-stress region along each notch is about 55.5 mm, falling within the region –27.5 mm ≤ x ≤ +27.5 mm. The uniformity of the stress in the constant-stress zone was found to be very good, to within a small fraction of one per cent.
As mentioned by Kaye (2010), the 5 mm buffer zone between the grip region and the end of the notch was chosen because a discontinuity occurs at the edge of the grip. The exact stress/strain state at this location is unknown, as it depends on the amount of surface yielding and/or slip that occurs over the grip area contact surface. Preliminary finite element analyses conducted by Wescott, Jones and Heller (2010) had also shown that excessive local stresses occur near the ends of the notch if a suitable buffer zone is not included.
A FADD2D ¼-symmetry 2D model of the Kaye (2010) coupon was created and analysed using the previously specified linear elastic material properties for aluminium. An initial FADD2D input deck was created using a custom-written Fortran 90 program called CreateDogboneCouponModel (see Appendix D for the source code listing of this program, and Appendix E for the associated input deck, both of which are also located in Objective folder fAV1044714). Note that the coordinates of the notch profile that were used here were those that had been obtained after the radius-of-curvature coordinate-smoothing process described by Wescott and Heller (2009) had been applied. The applied load once again consisted of a uniform uniaxial pressure of 100 MPa. The initial input deck was manually edited to convert it from a full model to the desired ¼-symmetry model.
The computed normalised stress response, σ/σmax, along the smoothed notch boundary is shown in Figure 8. The net-section SCF for this design is Kt = 1.038, which is in good agreement with the prior result of Kt = 1.030 that was reported by Kaye (2010), and the gross-section SCF is Ktg = 1.660. The normalised stress is quite constant over an arc length of approximately 54.4 mm along the notch in the central region of the coupon (–27.0 mm ≤ x ≤ +27.0 mm). Although the stress response is very flat, the stress level in this region is about 0.7% lower than the peak stress, which occurs at about x = 25.5 mm. This is likely to be a consequence of a number of interacting factors. Firstly, the smoothing process, which has been applied to produce the smoothed notch boundary, has a small detrimental impact on the uniformity of the boundary stresses, as originally noted by Kaye (2010) and subsequently further investigated by Evans, Yu and Heller (2015). Secondly, first-order
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 7
elements were used in the original Nastran-based shape optimisation, whereas the FADD2D code utilises 2nd-order boundary elements that are capable of more accurately resolving local stress variations associated with any given geometry. Finally, the optimal shape is based on a 3D finite element analysis, whereas the FADD2D analysis is 2D in nature.
4. Design of the modified coupon with optimal constant-stress notch profile
The prior constant-stress coupon design has experienced undesirable failures in the grip area during the fatigue test program, and a contributing factor is considered to be the use of oversized grips. By increasing the width of the coupon in the grip area, while maintaining the minimum width in the central region, it is expected that failures in the grip area will be avoided. As the previous fatigue testing work has already produced a quantity of results, it is required to create a new coupon design that maintains as much compatibility with those results as possible. In that respect, it is desired that the length of the constant-stress region in the modified coupon be very similar to that of the original design.
The previous MSC Patran/Nastran–based DST Group shape optimisation procedure has recently been modified for use with the Abaqus finite element analysis code by Kaye and Waldman (2016). This new version of the shape optimisation code was used here for designing the new Modified Constant-Stress Coupon. The Abaqus-based procedures are presently capable of performing a 2.5D optimisation analysis. This means that the y-displacement through the thickness of the model is maintained at each point on the notch profile while the profile is being optimised.
4.1 Nominal geometry and material properties
The nominal design of the Modified Constant-Stress Coupon is shown in Figure 9, and it is based on the previous design for a constant-stress coupon reported in Kaye (2010). As before, the new coupon is 6.35 mm thick, and has linear-elastic material properties typical of an aluminium alloy: E = 72.4 GPa and υ = 0.33. To improve the performance in the grip area, the gross width of the new coupon has been increased to 60 mm, with a gripped area that is now 60 mm × 60 mm at each end. The new design also includes a 25-mm buffer zone between the grip line and the end of the notched region, whereas it was 5 mm in the previous design described by Kaye (2010). The increased size of this buffer zone is expected to make the stresses less sensitive to different boundary conditions that might be utilised to simulate the grip loading. The total length of the new coupon is 244 mm. The minimum cross-sectional width in the notched region is the same as before (25 mm). As the notch depth in this Modified Constant-Stress Coupon is more severe than previously, the length of the notch region was increased from 70 mm to 74 mm. This will ostensibly
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 8
provide additional room over which the constant stress zone can develop, thus offsetting to some degree the effects of the more severe stress concentration geometry.
4.2 Initial finite element mesh
In order to perform the shape optimisation, it is necessary to create an initial finite element model. A ⅛-symmetry model of the coupon was developed in order to reduce computation time and to avoid potential numerical errors. The model was created using the MSC Patran pre- and post-processing software, and then an Abaqus input deck was written out. Figure 10 shows a general view of the 3D finite element mesh that was created. The model utilised six layers of 8-noded hexagonal elements through the half-thickness.
A side view of the finite element model with the optimisation zone highlighted is shown in Figure 11. The assumed starting shape of the notch is a straight line, which was subdivided into 80 segments of equal length. The grip load was modelled as a uniformly-distributed constant traction force, and it was implemented using nodal forces. For this linear-elastic model, the applied uniaxial traction load was equivalent to a force of 50.02 kN (calculated by summing up the individual contributions of the forces applied at the nodes). This leads to a gross-section average stress of 131.3 MPa, and a net-section average stress of 315.1 MPa.
As we will be performing a linear elastic stress analysis, the load level can be scaled to obtain suitable stress levels in the notch zone to suit the particular requirements of the fatigue test.
4.3 Performing the shape optimisation
Prior to performing the shape optimisation, some further processing of the initial Abaqus input deck was conducted in accordance with the method described by Kaye and Waldman (2016) in order to prepare it for use by the shape optimisation code. The entire notch region was chosen as the movable boundary during the optimisation process (see Figure 11). The nodal movement at each point along the notch boundary was calculated using the maximum through-thickness stress. A minimum radius of curvature constraint of 20 mm was applied. All seven nodes through the thickness at each location on the notch boundary were moved by the same amount at each iterative step. The optimisation process was allowed to proceed for 400 iterations, whereupon it was manually terminated.
4.4 Raw results of shape optimisation
The resulting raw shape for the optimised constant-stress coupon that was produced by the shape optimisation process is shown in Figure 12. The contours of maximum principal stress from the finite element analysis are shown in Figure 13. The distribution of normalised principal stress, σ/σmax, along the notch boundary is shown in Figure 14, where σ is the largest principal stress through the thickness at each location, and σmax is the
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 9
maximum value of the principal stress along the optimised notch boundary. The peak occurs at the centre of the notch boundary, and is σmax = 330.3 MPa. Using this value of stress, the gross-section SCF is Ktg = 2.516 for this configuration, while the net-section SCF is Kt = 1.048.
Referring to Figure 14, it is noted that, as expected, the normalised maximum principal stress in the central region of the optimised notch boundary is very uniform. This constant-stress zone extends over an arc length of approximately 51.2 mm, falling within the region –25.3 mm ≤ x ≤ +25.3 mm. The standard deviation of the normalised principal stress in the constant-stress region is only 0.006%, which is negligible, indicating the optimality of the solution. The arc length of the present constant-stress region is 7.7% less than that of the original constant-stress coupon.
In Figure 15, the maximum principal stress contour levels were adjusted to lie between a maximum of 330.3 MPa and a minimum of 310.5 MPa, with 12 contour bands in between. Each of these contour levels represents a 0.5% change in stress relative to the peak stress, for a total range of 6.0%. To provide some details of a higher resolution, Figure 16 shows contours of maximum principal stress that lie within 1.0% of the peak value.
As depicted in Figures 13–16, the distribution of the principal stress is very uniform along a large section of the notch surface of the optimised boundary contour, just as required. Furthermore, the variation of the maximum principal stress in the thickness direction is generally very low, this being particularly so at the centre of the coupon. Through-thickness variation in the maximum principal stress only becomes noticeable near the end of the constant stress zone. As expected, the through-the-thickness stress distribution peaks at the mid-plane of the coupon.
4.5 Preparation of raw optimal shape for manufacture
As described by Wescott and Heller (2009), and more recently by Evans, Yu and Heller (2015), past experience with the numerically-controlled machining of free-form notch shapes has indicated that it is desirable to process the notch coordinates so that the shape of the notch can be represented by a series of many short circular arcs. This conversion procedure also incorporates the facility for performing some smoothing of the raw notch shape in order to reduce the variability of the radius of curvature (ROC) along the notch boundary. In practice, the shape transformation process produces only very minor adjustments to the raw shape, and the resulting shape is still essentially free-form. In addition, the ROC-smoothing process has a very small effect on the stress distribution along the notch profile, which although noticeable is often negligible. Experience with some optimal shapes appears to indicate that the ROC-smoothing process can result in stresses that increase slightly near the end of the zone of constant stress (see Figure 8 for an example of this type of behaviour).
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 10
A plot of the ROC distribution computed from the nodal coordinates along the entire (mirrored) raw unsmoothed as-optimised notch boundary is shown in Figure 17. The ROC at any given node was computed by a fitting circular arc through the three sets of coordinates comprised of that node and its immediate left and right neighbours (except at the start and end nodes along the boundary). The effect of the ROC constraint of 20 mm is clearly visible at the two ends of the notch boundary. From the response displayed in Figure 17, it is apparent that there are significant fluctuations in the ROC, with a large number of very high ROC values being present. These large fluctuations in ROC value have been found to cause difficulties during the manufacturing of optimal shapes using standard computer numerically-controlled machining techniques.
The rapid and high-amplitude oscillations in the radius-of-curvature of the as-optimised geometry are potentially caused by the use of first-order elements during the iterative finite element analysis. Prior shape optimisation using second-order elements did not appear to display this rather unsatisfactory behaviour. Further investigation is warranted in order to remediate this problem. It is considered that it would be better not to apply the previously-developed automated smoothing procedures in a brute force manner. It would be preferable to determine the cause of the radius-of-curvature problems and address them directly.
Upon subsequent closer investigation here, these rapid and high-amplitude oscillations in the computed ROC values have been identified as being associated with very small transitions in slope between the edges of some of the linear 8-noded hexahedral first-order finite elements. These cause the large calculated ROC values associated with pairs of elements whose connected edges are nearly collinear. This behaviour is depicted in Figure 18, where the shape of the raw optimal profile near x = 0 mm has been plotted (solid line). The presence of these “flat” areas, with their commensurately-high ROC values (based on a 3-point circular arc curve fit), is clearly evident in the plot of the raw unsmoothed optimal profile. It is believed that these “flat” regions are caused by the numerical approximations inherent in the use of first-order finite elements to obtain the optimal solution, as prior shape optimisation solutions obtained using second-order finite elements did not display the type of oscillatory ROC behaviour found here (see Waldman, Heller and Chen 2001; Waldman and Heller 2006; Waldman and Heller 2015). The dashed line shows the results of fitting a smooth curve through every second point of the original data set, resulting in a smooth line that accurately follows the underlying shape of the optimal profile.
Hence, as in the prior work performed by Kaye (2010), it is necessary here to undertake ROC smoothing of the raw geometric coordinates of the optimised boundary shape in order to make the notch suitable for manufacturing. Based on an assessment of the relevant information presented in Figure 17 and Figure 18, it was decided to accomplish this smoothing by using a two-phase process in order to produce a result of sufficiently high quality.
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 11
In the first phase of the smoothing process, the raw optimal shape was curve fit using Microsoft Excel, with the problem coordinates that produced the high-amplitude ROC values omitted from the fit. Those points were then manually modified to lie closer to the interpolated shape, thus greatly reducing the severity of the local ROC variations. The result of this process is depicted in Figure 19. There it is evident that the extreme ROC variations have been reduced quite significantly, producing a much smoother overall behaviour. This result is now much more in keeping with our expectations for an optimal shape, but some further smoothing of the ROC variations is still desirable.
In the second phase of the radius of curvature smoothing process, the techniques reported by Wescott and Heller (2009) were applied to further smooth the notch shape. Two cycles of this automated smoothing were computed, and the ROC results obtained are depicted in Figure 20. The (x, y) coordinates at the two ends of the notch, as well as the (x, y) coordinate associated with the centre of the notch shape, were held fixed during the ROC smoothing process. Because of the manual adjustment of the node coordinates prior to undertaking the next phase of ROC smoothing, the automated ROC smoothing process has been very effective in further reducing the variations in ROC along the notch boundary, even with just two cycles.
Note that this level of ROC smoothing has a very small effect on the stresses in the constant-stress zone. This was checked by creating a new finite element model based on the smoothed coordinates of the optimal boundary. The normalised maximum principal stress for the smoothed notch geometry is shown in Figure 21. Compared to the results for the unsmoothed shape, the corner at approximately x = 25 mm is slightly more rounded, and the extent of the constant stress zone has been reduced a little bit (from x = 25.3 mm to x = 24.8 mm). The arc length of the constant-stress zone for the smoothed notch is 50.3 mm, and this is 1.8% shorter than for the unsmoothed notch.
The smoothed notch coordinates are supplied in Table 1. These were read into MSC Patran and used to generate an IGES file containing a representation of the smoothed optimal notch profile suitable for use in creating commands for computer numerically controlled machining of the coupons. The IGES file format is used to specify the notch boundary as a series of circular arcs and straight line segments. A listing of the contents of the IGES file corresponding to the notch contour for the Modified Constant-Stress Coupon is provided in Appendix G (a copy of this file can also be found in Objective folder fAV1044714, and the name of the file is Al_mod_optimal_3D_msmth_smoothed_002.igs).
It is instructive to compare the smoothing approach used above with the results that are obtained when the original raw unsmoothed coordinates for the Modified Constant-Stress Coupon have been passed through 20 cycles of the automated ROC smoothing algorithm. The resulting ROC variation in the notch region is shown in Figure 22. It is apparent that these results are clearly different from those presented in Figure 20, with the major differences being confined to the central region of the notch, –7 mm ≤ x ≤ +7 mm. The ROC results obtained using 20 cycles of automated smoothing have a peak ROC value that is
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 12
about 15% greater. It is considered here that the application of manual plus automated smoothing has produced a smoother overall result, and it is expected that the resulting changes to the shape coordinates will be less than those produced by 20 cycles of automated smoothing.
5. 2D fracture mechanics assessment of modified constant-stress coupon
As the Modified Constant-Stress Coupon will be used in fatigue testing studies involving crack growth, it is instructive to assess its performance from a fracture mechanics perspective. The programs that were used for this analysis are located in Objective folder fAV1044714.
5.1 Analysis of uncracked coupon
Firstly, a full 2D model of the uncracked Modified Constant-Stress Coupon was created for analysis with FADD2D using the coordinates of the node locations in the original finite element model. A full model was needed as this forms the basis for subsequent stress intensity factor calculations associated with the placement of through-thickness edge cracks at various assumed locations distributed along the notch boundary.
The material properties corresponded to those for a typical aluminium alloy, and were the same as before (E = 72.4 GPa and υ = 0.33). Plane stress conditions were specified as the thickness of the specimen is small compared to the other dimensions. The applied load consisted of a uniform tensile pressure applied to the vertical edges at the two ends. To prevent rotation of the grip ends, 6 points were chosen on the grip zone boundary and restrained from moving in the y-direction. Using the (x, y) coordinate system defined in Figure 9, these points were located at (x, y) = (±62 mm, 20 mm), (±62 mm, –20 mm) and (±122 mm, 0 mm). In addition, in order to prevent rigid body motion of the coupon, the point on the boundary at (x, y) = (–122 mm, 0 mm) was restrained from moving in the x-direction.
The FADD2D model was created using the CreateDogboneCouponModel program (see Appendix D), and it was used to compute the maximum principal stress along the optimised notch boundary of the modified constant stress coupon. Figure 23 plots the normalised maximum principal stress obtained from the boundary element model, and compares it to the results obtained from the finite element model that was used during the shape optimisation process. It is evident that the two quite different computational approaches have produced results that are very similar. Using the FADD2D boundary element method, the gross-section SCF is Ktg = 2.519, which is almost identical to the value Ktg = 2.516 obtained using the finite element method.
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 13
5.2 Computation of crack trajectories for edge-cracked coupon
Using the 2D model described above, an edge crack was inserted at various locations along the notch boundary. As a result of the 2D nature of the solution, the edge crack is assumed to be a through crack. In each case, the initial crack length set to be 0.1 mm, which can be regarded as being a very short crack relative to the size of the specimen. At each chosen location, the simulated crack was oriented to be perpendicular to the boundary. This essentially means that the crack was initially aligned to open in a Mode I dominant crack propagation mode, as this was assumed to be representative of what might happen in the coupon during fatigue testing.
For the purposes of this analysis, the cracks are located at different points along the notch boundary with the following x-coordinates: 0.00 mm, 4.24 mm, 8.51 mm, 12.74 mm, 16.95 mm, 21.11 mm, 23.17 mm, 25.26 mm, 29.45 mm, and 32.14 mm. These starting points correspond to node locations in the original finite element model. The FADD2D code was used to compute the crack trajectory for a single crack growing from each of these locations. A crack growth increment of 0.1 mm per iteration was used to compute the crack growth trajectory associated with each cracking scenario. For the purpose of this simulation, the cracks were allowed to grow until they had reached 20 mm in length.
The computed crack trajectories obtained for each of the cracks are shown in Figure 24. The trajectories become more curved as the starting location of the crack moves further away from the centre of the coupon. The crack trajectories remain relatively straight in the range of starting locations 0 mm ≤ x ≤ 15 mm.
5.3 Computation of Beta factors for through-thickness edge-cracked coupon
When the crack trajectories were being computed, the FADD2D boundary element code was also calculating the corresponding stress intensity factors, K, for the particular crack geometry that was present at each crack growth increment. It is considered instructive to compare these stress intensity factors to those corresponding to an edge crack of equivalent length that is located in a semi-infinite plate.
For a through-thickness edge crack of length a in a semi-infinite plate subjected to a remote uniaxial tension stress, S∞, the equation for the stress intensity factor K is
FaSK π= ∞ (1)
where F is a boundary correction factor (also called a Beta factor). As its name implies, the boundary correction factor accounts for the influence of various boundaries, but it is also often a function of parameters such as crack geometry, crack length, plate width, plate thickness, hole radius (where present), angular position along a crack front, etc. However, for the semi-infinite plate geometry, the boundary correction factor is a constant value,
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 14
and is simply F = 1.1215 (Tada, Paris and Irwin 2000; Broek 1978), which is also known as the free-edge correction factor.
For a through-thickness edge crack that is located at any point along the notch boundary of the Modified Constant-Stress Coupon, the stress intensity factor can be written as
FaSK π= (2)
where S is the local tangential stress at the notch surface, and a is the crack length as measured along the curved crack trajectory. In the constant stress region, S can be computed by using S = Ktg S∞, where Ktg is the local gross-section SCF and S∞ is the remote uniform uniaxial tension stress applied to the ends of the coupon.
Figure 25 shows the boundary correction factors, F, plotted as a function of crack length, a, for a number of cracks starting from different points along the notch boundary. As a point of reference, the crack-length-independent boundary correction factor for a through-thickness edge crack in a semi-infinite plate is represented by the dashed horizontal line (F = 1.1215). For cracks located in the range 0 mm ≤ x ≤ 16.95 mm that have crack lengths a ≤ 2.5 mm, it is apparent that the geometry factor is within about 2% or so of F = 1.1215 for an edge crack in a semi-infinite plate.
The boundary correction factors as a function of distance along the notch boundary for edge cracks of varying lengths for the Modified Constant-Stress Coupon are plotted in Figure 26. The boundary correction factor for very short cracks, a ≤ 0.1 mm, is within 1.0% of the theoretical edge crack solution, F = 1.1215, over the entire length of the constant stress region. For short cracks, a ≤ 1.0 mm, the boundary correction factor is within 3.3% of F = 1.1215 along 84% of the constant-stress region, dropping by 7.5% at the end of the constant-stress zone.
5.4 Computation of Beta factors for other crack geometries
From the data presented in Figure 25 and Figure 26, the stress intensity factors for short through-thickness edge cracks for the constant stress specimen are largely in agreement with the solution for an edge crack in a semi-infinite plate. By analogy, it would therefore be expected that stress intensity factors for other types of short cracks (e.g. circular or elliptical corner cracks, circular or elliptical surface cracks, etc.) would be well represented by their nominal solutions computed for geometries that are similar to that of the constant-stress coupon. For example, if so desired, the stress intensity factor equations for planar semi-elliptical surface and corner flaws published by Broek could be utilised (Broek 1978, Section 3.6, Elliptical cracks). Hence, for many types of small (short) natural flaws, the body of existing stress intensity factor solutions can be taken advantage of, thus circumventing the need for computation of new sets of solutions that are specific to the present coupon geometry and loading.
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 15
6. Conclusion The coordinates of the smoothed notch profile for the new Modified Constant-Stress Coupon were presented in Table 1. These have also been made available as an IGES file, which can be utilised for manufacturing of the coupons using numerically-controlled machining equipment. A listing of the contents of the IGES file is given in Appendix G, and a copy of the file can be found in Objective folder fAV1044714 (the name of the file is Al_mod_optimal_3D_msmth_smoothed_002.igs).
A summary of the dimensions of the Modified Constant-Stress Coupon, together with those of the original design, is presented in Table 2. It is noted that the grip area of the Modified Constant-Stress Coupon has been increased by 225% over the original constant-stress coupon design. Together with the increase in gross-section SCF resulting from the increase in the total width of the coupon, the incidences of specimen failure in the grip zone are expected to be greatly reduced, if not entirely eliminated. The arc length of the constant-stress zone in the Modified Constant-Stress Coupon is approximately 7.5% less than that which was obtained for the original constant-stress coupon. If required, a closer match could be attained by increasing the length of the notch zone and/or reducing the depth of the notch.
Using a 2D boundary element model of the Modified Constant-Stress Coupon, a number of potential crack growth trajectories have been computed by using the FADD2D fracture analysis code to perform simulations of 2D through-thickness edge cracks. The starting locations of the edge cracks were distributed along the constant-stress zone of the notch boundary. As the cracking location moved away from the centre of the coupon, the computed crack trajectories progressively became more curved. In the event that cracks that develop in the coupons are allowed to grow to a long length, it would be interesting to compare the crack growth trajectories obtained from fatigue testing with their computationally-determined counterparts.
7. Acknowledgement The author would like to thank Professor Mark E Mear, University of Texas at Austin, and Professor James C Newman Jr, Mississippi State University, for providing access to the FADD2D boundary element analysis code that was utilised for the fracture mechanics calculations in this report.
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 16
8. References ASTM E606/E606M-12, Standard Test Method for Strain-Controlled Fatigue Testing, ASTM International, West Conshohocken, PA, 2012, www.astm.org.
Barter SA, L Molent, RJH Wanhill. Typical fatigue-initiating discontinuities in metallic aircraft structures. International Journal of Fatigue, Vol 41, 2012, pp 11–22.
Braemar R. Code enhancements for the Patran/Nastran structural optimisation. DSTO Minute, File B2/129/Pt4, Melbourne, 20 May 2005.
Broek D. Elementary Engineering Fracture Mechanics, Second Edition. Sitjhoff & Noordhoff, The Netherlands, 1978.
Burchill M, M Heller. Optimal notch shapes for loaded plates. Journal of Strain Analysis for Engineering Design, Vol 39, No 1, 2004, pp 99–116.
Cetin A, G Härkegård, A Naess. The fatigue limit: An analytical solution to a Monte Carlo problem. International Journal of Fatigue, Vol 55, 2013, pp 194–201.
Chang C, ME Mear. A boundary element method for two dimensional linear elastic fracture analysis. International Journal of Fracture, Vol 74, 1995, pages 219–251.
Evans R, X Yu, M Heller. Transfer effects for stress optimal shapes between design codes and from design to NC manufacture. 11th World Congress on Structural and Multidisciplinary Optimisation, 7–12 June, 2015, Sydney Australia.
Heller M, M Burchill, R Wescott, W Waldman, R Kaye, R Evans, M McDonald. Airframe life extension by optimised shape reworking – Overview of DSTO developments. 25th ICAF Symposium – Rotterdam, 27–29 May 2009.
Heller M, R Kaye, LRF Rose. A gradientless finite element procedure for shape optimization. Journal of Strain Analysis, Vol 34, No 5, 1999, pp 323–336.
Kaye R. Follow-on analysis of uniform stress coupons for combat aircraft life assessment. DSTO Minute, File B2/129/Pt4, 16 December 2010.
Kaye R, W Waldman. Conversion of DST Group shape optimisation software for increased portability across computing platforms. DST Group Technical Report, 2016 (DST-Group-TR-3251).
Loader C, A Shekhter, S Turk. Assessment of the effect of anodizing on fatigue life of aluminium alloys using equivalent crack size modelling methods. Aging Airworthiness Aircraft and Sustainment Conference, Brisbane, 2015.
Loader C, A Shekhter, S Turk, J Niclis, K Sharp. Investigation into the effect of sulphuric acid anodising on the fatigue life of AA7050-T7451. DST Group Technical Report, 2016 (to be published).
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 17
Mattheck C, D Erb, K Bethge, U Begemann. Three-dimensional shape optimisation of a bar with a rectangular hole. Fatigue & Fracture of Engineering Materials & Structures, Vol 15, Issue 4, April 1992, pp 347–351.
McDonald M. FEA-COUPON-001 (OPT), Rev 2, Coupons Stress Analysis. DSTO Internal Document, 12 June 2009a.
McDonald M. FEA-COUPON-002 (OPT), Rev 1, Optimal shapes for coupons. DSTO Internal Document, 9 June 2009b.
Molent L, SA Barter, RJH Wanhill. The lead crack fatigue lifing framework. International Journal of Fatigue, Vol 33, Issue 3, 2011, pp 323–331.
Newman Jr JC, C Chang, L Xiao, ME Mear, VJ Kale. FADD2D: Fracture Analysis by Distributed Dislocations, Version 1.0, User Guide for Personal Computers with Demonstration Example. October 2006.
Niclis J, T Harrison. Fatigue Implications of surface treated AA7050-T7451 tested under VA loading. DST Group Technical Report, 2016 (to be published).
Shekhter A, C Loader, S Turk, PK Sharp. Effect of anodising treatments on equivalent crack size of the 7XXX aluminium alloy. 16th Australian International Aerospace Congress, 23–24 February 2015, Melbourne, Australia.
Shekhter A, C Loader, S Turk, J Niclis. Fatigue implications of surface treated AA7085-T7452 tested under CA and VA loading. DST Group Technical Report, 2016 (to be published).
Tada H, PC Paris, GR Irwin. The Stress Analysis of Cracks Handbook. Third Edition, Professional Engineering Publishing, 2000.
Turk S, J Niclis. Microstructural and fractographic assessment of fatigue crack behaviour of surface treated AA7085-T7452. DST Group Technical Report, 2016 (to be published).
Turk S. Fatigue Implications of sulphuric acid anodising on AA7050-T7451 aluminium alloy. 10th International Conference on Structural Integrity and Failure (SIF-2016): Advances in Materials and Structures, 12–15 July 2016a, Adelaide, Australia (to be presented).
Turk S. Fractographic analysis and fatigue crack growth of surface treated AA7050-T7451 coupons. DST Group Technical Report, 2016b (to be published).
Waldman W, M Heller, GX Chen. Optimal free-form shapes for shoulder fillets in flat plates under tension and bending. International Journal of Fatigue, Vol 23, 2001, pp 509–523.
Waldman W, M Heller. Shape optimisation of holes for multi-peak stress minimisation. Australian Journal of Mechanical Engineering, Vol 3, No 1, 2006, pp 61–71.
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 18
Waldman W, M Heller. Shape optimisation of holes in loaded plates by minimisation of multiple stress peaks. DSTO Research Report DSTO-RR-0412, April 2015.
Waldman W. Determination of minimised Kt values and boundary shapes for a class of quasi-rectangular holes in infinite plates. DST Group Technical Report TR-3125, July 2015.
Wescott R, M Heller. Transforming stress optimal free form shapes for improved numerically controlled manufacture. DSTO Research Report DSTO-RR-0340, July 2009.
Wescott R, M Jones, M Heller. Stress analysis for design of uniform stress coupons for combat aircraft life assessment. DSTO Minute, File B2/129/Pt4, 14 July 2010.
Yu X, M Burchill, R Kaye, S Barter. Optimal coupon design to achieve natural crack start in coupon fatigue tests. In: Recent Advances in Structural Integrity Analysis: Proceedings of the International Congress (APCF/SIF-2014). Editors: L Ye, A Kotousov, L Chang. Darlington Campus, University of Sydney, Australia, 9–12 December 2014, pp 137–141.
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 19
Table 1: Coordinates of the smoothed optimal profile for Modified Constant-Stress Coupon. (See Figure 9 for definition of the x-y coordinate system.)
write(luo,'(a)') '' if (cracked) then call writedashedline(luo,ndash) write(luo,'(a)') 'Definition of Crack' write(luo,'(a)') '' nelemcrack = 8 do i = 1,ncracks write(luo,'(i4.4,i2)') i,1 write(luo,'(a)') ' 1 0 0 2 1' j = icrackpoint(i) if (mod(j,2) .eq. 0) then write(*,*) write(*,*) 'Error: Crack location index must be an odd number. Check your input data.' write(*,*) stop endif dx = a0*xinwnv(j) dy = a0*yinwnv(j) write(luo,'(2i4,4f18.7,i6,f18.3)') 1,1,cx(j),cy(j),cx(j)+dx,cy(j)+dy,nelemcrack,anglenormal(j) write(luo,'(a)') '' enddo endif call writedashedline(luo,ndash) write(luo,'(a)') 'Definition of Boundary' write(luo,'(a)') '' write(luo,'(a)') '1 1' write(luo,'(i4.4,i2,i6)') ncracks,0,4*(nc/2)+nb-2 ! Initialise counter that gets incremented when each ! boundary segment is created. id = 0 ! Write out the LH upper quadrant of the curved boundary. nelem = 2 do i = 1,nc/2 ic = (i-1)*2+1 if (i.eq.1 .and. cn(1).gt.0) then call segpara(-cx(ic),+cy(ic),-cx(ic+1),+cy(ic+1),-cx(ic+2),+cy(ic+2), & cn(ic),0,0,luo,id,8*nelem) else call segpara(-cx(ic),+cy(ic),-cx(ic+1),+cy(ic+1),-cx(ic+2),+cy(ic+2), & 0,0,0,luo,id,nelem) endif enddo ! Write out next line segments. nelem = 4 call segline(bx(01),by(01),bx(02),by(02),luo,id,nelem,-1,0.0d0,-1,0.0d0,3,0,1) call segline(bx(02),by(02),bx(03),by(03),luo,id,nelem,-1,0.0d0,-1,0.0d0,1,0,1)
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 44
call segline(bx(03),by(03),bx(04),by(04),luo,id,nelem,-1, -px ,-1,0.0d0,3,1,1) call segline(bx(04),by(04),bx(05),by(05),luo,id,nelem,-1, -px ,-1,0.0d0,1,1,1) call segline(bx(05),by(05),bx(06),by(06),luo,id,nelem,-1,0.0d0,-1,0.0d0,3,0,1) call segline(bx(06),by(06),bx(07),by(07),luo,id,nelem,-1,0.0d0,-1,0.0d0,1,0,1) ! Write out the LH lower quadrant of the curved boundary. nelem = 1 do i = 1,nc/2 ic = nc-(i-1)*2 call segpara(-cx(ic),-cy(ic),-cx(ic-1),-cy(ic-1),-cx(ic-2),-cy(ic-2), & 0,0,0,luo,id,nelem) enddo ! Write out the RH lower quadrant of the curved boundary. nelem = 1 do i = 1,nc/2 ic = (i-1)*2+1 call segpara(+cx(ic),-cy(ic),+cx(ic+1),-cy(ic+1),+cx(ic+2),-cy(ic+2), & 0,0,0,luo,id,nelem) enddo ! Write out next line segments. nelem = 4 call segline(bx(08),by(08),bx(09),by(09),luo,id,nelem,-1,0.0d0,-1,0.0d0,3,0,1) call segline(bx(09),by(09),bx(10),by(10),luo,id,nelem,-1,0.0d0,-1,0.0d0,1,0,1) call segline(bx(10),by(10),bx(11),by(11),luo,id,nelem,-1, +px ,-1,0.0d0,3,0,1) call segline(bx(11),by(11),bx(12),by(12),luo,id,nelem,-1, +px ,-1,0.0d0,1,0,1) call segline(bx(12),by(12),bx(13),by(13),luo,id,nelem,-1,0.0d0,-1,0.0d0,3,0,1) call segline(bx(13),by(13),bx(14),by(14),luo,id,nelem,-1,0.0d0,-1,0.0d0,1,0,1) ! Write out the RH upper quadrant of the curved boundary. nelem = 2 do i = 1,nc/2 ic = nc-(i-1)*2 ! If a crack is located at either the start or end node of the ! parabolic segment, increase the mesh density for the segment. if (cn(ic).gt.0 .or. cn(ic-2).gt.0) then call segpara(+cx(ic),+cy(ic),+cx(ic-1),+cy(ic-1),+cx(ic-2),+cy(ic-2), & cn(ic),cn(ic-1),cn(ic-2),luo,id,8*nelem) else call segpara(+cx(ic),+cy(ic),+cx(ic-1),+cy(ic-1),+cx(ic-2),+cy(ic-2), & cn(ic),cn(ic-1),cn(ic-2),luo,id,nelem) endif enddo write(luo,'(a)') ' ' call writedashedline(luo,ndash) close(luo) end subroutine DogboneCouponModel
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 45
!============================================================================= subroutine writedashedline(luo,ndash) implicit none integer(4) luo,ndash,i write(luo,'(2000a)') ('-',i=1,ndash) end subroutine writedashedline !============================================================================= subroutine segpara(x1,y1,x2,y2,x3,y3,cn1,cn2,cn3,luo,id,nelem) implicit none real(8) x1,y1,x2,y2,x3,y3 integer(4) cn1,cn2,cn3 integer(4) luo,id,nelem id = id + 1 write(luo,'(a)') '' write(luo,'(i4.4,2i4)') id,1,nelem write(luo,'(2f15.7,i4,f15.7,i4,f15.7,i4)') x1,y1,-1,0.0d0,-1,0.0d0,cn1 write(luo,'(2f15.7,i4,f15.7,i4,f15.7,i4)') x2,y2,-1,0.0d0,-1,0.0d0,cn2 write(luo,'(2f15.7,i4,f15.7,i4,f15.7,i4)') x3,y3,-1,0.0d0,-1,0.0d0,cn3 return end subroutine segpara !============================================================================= subroutine segline(x1,y1,x2,y2,luo,id,nelem,icx,vcx,icy,vcy,icn,icnx,icny) implicit none real(8) x1,y1,x2,y2,vcx,vcy integer(4) luo,id,nelem,icx,icy,icn,icnx,icny real(8) px real(8) xm,ym id = id + 1 xm = (x1+x2)/2.0d0 ym = (y1+y2)/2.0d0 write(luo,'(a)') '' write(luo,'(i4.4,2i4)') id,1,nelem if (icn.eq.0) then write(luo,100) x1,y1,icx,vcx,icy,vcy,0 write(luo,100) xm,ym,icx,vcx,icy,vcy,0 write(luo,100) x2,y2,icx,vcx,icy,vcy,0
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 46
else if (icn.eq.1) then write(luo,100) x1,y1,icx,vcx,icy,vcy,-2,icnx,icny else write(luo,100) x1,y1,icx,vcx,icy,vcy,0 end if if (icn.eq.2) then write(luo,100) xm,ym,icx,vcx,icy,vcy,-2,icnx,icny else write(luo,100) xm,ym,icx,vcx,icy,vcy,0 end if if (icn.eq.3) then write(luo,100) x2,y2,icx,vcx,icy,vcy,-2,icnx,icny else write(luo,100) x2,y2,icx,vcx,icy,vcy,0 end if end if return 100 format(2f15.7,i4,f15.7,i4,f15.7,3i4) end subroutine segline !============================================================================= subroutine anglevector(n,xc,yc,angletangent,anglenormal,xinwnv,yinwnv) ! Determine the tangent angle, inward normal angle, and inward normal vector ! for points along the curved optimal boundary. It is assumed that the first ! point is mirrored about the y-axis. implicit none integer(4) n real(8) xc(n),yc(n),angletangent(n),anglenormal(n),xinwnv(n),yinwnv(n) integer(4) i real(8) pi,pion2,twopi pi = 4.0d0*atan(1.0d0) pion2 = pi/2.0d0 twopi = 2.0d0*pi do i=1,n ! Compute angle of slope of the curve at each point on the curve. if (i.eq.1) then angletangent(i)=atan2(0.0d0,xc(i+1)-(-xc(i+1))) else if (i.eq.n) then angletangent(i)=atan2(yc(i)-yc(i-1),xc(i)-xc(i-1)) else ! This simple formula for calculating the slope of the ! curve is equivalent to what is obtained by using ! a second-order Lagrange polynomial approximation. angletangent(i)=atan2(yc(i+1)-yc(i-1),xc(i+1)-xc(i-1)) endif ! Compute angle of the outward normal.
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 47
anglenormal(i)=angletangent(i)-pion2 if (anglenormal(i).gt.pi) anglenormal(i)=anglenormal(i)-twopi ! Compute inward unit normal vector to boundary. xinwnv(i)=cos(anglenormal(i)) yinwnv(i)=sin(anglenormal(i)) ! Convert angles from radians to degrees. anglenormal(i)=anglenormal(i)*180.0d0/pi angletangent(i)=angletangent(i)*180.0d0/pi enddo return end !============================================================================= subroutine circle(x1,y1,x2,y2,x3,y3,xc,yc,r,det) ! Determine the centre coordinates (xc,yc) and radius r of a circle passing ! through three points, (x1,y1) (x2,y2) (x3,y3). ! ! If the value of det is zero, then a circle through the three given points ! does not exist. e.g. the three points are colinear. implicit none real*8 x1,y1,x2,y2,x3,y3,xc,yc,r,det real*8 a1,a2,c1,c2,b1,b2 a1=2.0*(x2-x1) a2=2.0*(x3-x2) b1=2.0*(y2-y1) b2=2.0*(y3-y2) c1=x2**2+y2**2-x1**2-y1**2 c2=x3**2+y3**2-x2**2-y2**2 det=a1*b2-b1*a2 if (det.ne.0.0) then xc=(c1*b2-b1*c2)/det yc=(a1*c2-c1*a2)/det a1=xc-x1 a2=yc-y1 r=dsqrt(a1**2+a2**2) else xc=0.0d0 yc=0.0d0 r=0.0d0 endif return end !============================================================================= subroutine canglevector(n,xc,yc,angletangent,anglenormal,xinwnv,yinwnv)
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 48
! Determine the tangent angle, inward normal angle, and inward normal vector ! for points along the curved optimal boundary. It is assumed that the first ! point is mirrored about the y-axis. implicit none integer(4) n real(8) xc(n),yc(n),angletangent(n),anglenormal(n),xinwnv(n),yinwnv(n) integer(4) i real(8) pi,pion2,twopi,r,cx,cy,det pi = 4.0d0*atan(1.0d0) pion2 = pi/2.0d0 twopi = 2.0d0*pi do i=1,n ! Compute angle of slope of the curve at each point on the curve. if (i.eq.1) then call circle(-xc(i+1),yc(i+1),xc(i),yc(i),xc(i+1),yc(i+1),cx,cy,r,det) if (det.eq.0.0d0) then angletangent(i)=atan2(0.0d0,xc(i+1)-(-xc(i+1))) endif else if (i.eq.n) then call circle(xc(i-2),yc(i-2),xc(i-1),yc(i-1),xc(i),yc(i),cx,cy,r,det) if (det.eq.0.0d0) then angletangent(i)=atan2(yc(i)-yc(i-1),xc(i)-xc(i-1)) endif else call circle(xc(i-1),yc(i-1),xc(i),yc(i),xc(i+1),yc(i+1),cx,cy,r,det) if (det.eq.0.0d0) then angletangent(i)=atan2(yc(i+1)-yc(i-1),xc(i+1)-xc(i-1)) endif endif ! Compute angle of the outward normal. if (det.eq.0.0d0) then anglenormal(i)=angletangent(i)-pion2 else anglenormal(i) = atan2(yc(i)-cy,xc(i)-cx) angletangent(i) = anglenormal(i)+pion2 endif if (anglenormal(i).gt.pi) anglenormal(i)=anglenormal(i)-twopi ! Compute inward unit normal vector to boundary. xinwnv(i)=cos(anglenormal(i)) yinwnv(i)=sin(anglenormal(i)) ! Convert angles from radians to degrees. anglenormal(i)=anglenormal(i)*180.0d0/pi angletangent(i)=angletangent(i)*180.0d0/pi enddo return end
UNCLASSIFIED DST-Group-TR-3252
UNCLASSIFIED 49
Appendix E:
Input deck used to create FADD2D model of original constant-stress coupon using
CreateDogboneCouponModel program
The following input file can also be found stored in Objective folder fAV1044714.
2. TITLE A Modified Constant-Stress Coupon for Enhanced Natural Crack Start during Fatigue Testing
3. SECURITY CLASSIFICATION (FOR UNCLASSIFIED REPORTS THAT ARE LIMITED RELEASE USE (L) NEXT TO DOCUMENT CLASSIFICATION) Document (U) Title (U) Abstract (U)
4. AUTHOR(S) Witold Waldman, Robert Kaye and Xiaobo Yu
5. CORPORATE AUTHOR Defence Science and Technology Group 506 Lorimer St Fishermans Bend Victoria 3207 Australia
6a. DST GROUP NUMBER DST-Group-TR-3252
6b. AR NUMBER AR-016-590
6c. TYPE OF REPORT Technical Report
7. DOCUMENT DATE May 2016
8. OBJECTIVE FOLDER ID fAV1044714
9. TASK NUMBER AIR 07/283
10. TASK SPONSOR OIC-ASI-DGTA
11. NO. OF PAGES 95
12. NO. OF REFERENCES 33
13. DST GROUP PUBLICATIONS REPOSITORY http://dspace.dsto.defence.gov.au/dspace/
14. RELEASE AUTHORITY Chief, Aerospace Division
15. SECONDARY RELEASE STATEMENT OF THIS DOCUMENT
Approved for public release OVERSEAS ENQUIRIES OUTSIDE STATED LIMITATIONS SHOULD BE REFERRED THROUGH DOCUMENT EXCHANGE, PO BOX 1500, EDINBURGH, SA 5111 16. DELIBERATE ANNOUNCEMENT No Limitations 17. CITATION IN OTHER DOCUMENTS Yes 18. RESEARCH LIBRARY THESAURUS Shape optimisation, Stress concentration, Fatigue testing, Constant-stress coupon, Fatigue coupon, Stress intensity factor, Fillet, Notch analysis, Finite element analysis, Boundary element analysis, Natural crack start, Beta factors 19. ABSTRACT This report details the development of a modified constant-stress coupon for use in fatigue testing. This novel coupon design has a significantly greater surface area along the notch boundary that is subjected to the peak stress, and it is useful in studies of the probabilistic growth of in-service fatigue cracks where fatigue life is governed by the most severe defect. The extended region of uniform stress is achieved by shape optimisation of the notch boundary, resulting in the elimination of the highly-localised stress concentration that is a characteristic feature of a traditional dog-bone coupon. The presence of an extensive region of uniform stress increases the incidence of fatigue cracking from small naturally-occurring surface imperfections or discontinuities, as well as more uniformly distributing the fatigue cracking over the region of constant stress. Stress intensity factors have also been computed for simulated crack growth trajectories for edge cracks starting at various locations distributed along the notch boundary. Use of the constant-stress coupon reduces the number of coupons that need to be tested in order to attain desired statistical confidence levels, leading to significant time savings and greatly reduced costs in conducting fatigue testing programs studying the initiation and growth behaviour of small cracks.